Every trade since 1999 that we can reconstruct in full, graded twice — on what was knowable beforehand and on how it turned out. A trade can be a sound decision that lost, or a poor one that won, and the two columns separate them. Click any row to open it in the Trade Lab.
| # | Trade | Teams | Date | Process | Likely outcome | Outcome |
|---|
Reading these. Sorted by Process or Outcome, each trade shows the side that column is about, named in the Teams column: descending for the winner, ascending for the loser. Sorted by Date, every side is listed together under its trade; sorted by Teams, every side is its own line, so a club's deals sit together. Process prices the trade on ratings that existed before the season; Outcome prices it on what the contracts actually returned. Likely outcome draws where trades graded like this one landed, by process value and size, the way the draft board draws a prospect: the shape is the spread from the 5th to the 95th percentile, the amber tick the middle outcome, the dashed line break-even, the diamond this trade's actual outcome (fitted on 1999–2025 trades; its 80% range held 80% of outcomes when checked forward in time). Each row has its own scale, so read a shape against its break-even line rather than against the rows beside it; hover for the numbers. An outcome outside the 80% range is marked ◆ in the Outcome column. Its middle sits well inside the process value, because outcomes move only 20 to 40 cents per dollar of process. Both run over each player's whole career: his remaining contract at the trade, then re-signing rights to age 40 — a chain of four-season deals, each at the chance he re-signs with the club holding him and the chance the deal is a max deal, times his value above the max (negative if he falls below it; a non-max deal pays him his value), the draft board's own rule (in Outcome, every season he actually went on to play there). The club that sends him away is charged the same amount; that assumes the sender would have re-signed him too, which is not true of a player demanding a trade. Untick charge senders for re-signing rights to keep the acquirer's credit but not charge the sender, in which case a trade no longer nets to zero. Both use the same constants -- that era's aging curve and win calibration, and one market price of a win (fit on every season's signings) scaled by that season's cap -- so the difference between them is what the players did rather than how the model was calibrated. Draft picks are priced by the forward model at their original slot year; protections are recorded but not their thresholds, so a protected pick is priced as unprotected and flatters whoever received it.
† Before 2003 the process figure uses the published aging curve and win calibration: the seasons before are too few to refit them. Ratings are still as-of, so what those rows know with hindsight is how aging and wins convert -- not who was good, which is the hindsight a forecast has to be denied.